Mean Reversion Trading Systems and Cryptocurrency Trading [A Deep Dive]

Prices move in a wave like fashion, moving back and forth following a broader trend. While doing so, it often revolves around a mean. It might move across or bounce off the mean. Mean reversion systems are designed to exploit this tendency.

Prices much above the mean are considered expensive and the ones below the mean are considered cheap and the positions are taken accordingly. This back and forth movement of prices or the standard deviation needs to be high enough to make profits otherwise the system might fail. There would not be much room to take profits if the volatility is very low. Costs such as slippage, brokerage etc. will make it difficult to remain profitable.

We will first look at frequency distribution of prices of Bitcoin and
visualize the mean reversal tendency through curves. Then we will move
on to create a simple counter-trend strategy for the prevailing trend in
the crypto space.

Statistical Evidence for Mean Reversion :

Rather than using frequency distribution of prices itself, we have
calculated frequency distribution of hourly returns. Lets look at some
of the statistics that I obtained after analyzing hourly price returns
for BTC/USDC pair trading on Coinbase from 13th November 2018 to 5th
April 2019 and BTC/USD pair trading on Coinbase from July 2017 to April
2019.

It can been seen that the graphs are not perfectly Bell Shaped. It is
worth noting that the values are very tightly distributed and it
resembles a double exponential distribution.With a little tolerance
mean, median and mode are almost equal.

Quoting mathworld.wolfarm.com :

“Reversion to the mean, also called regression to the mean, is the statistical phenomenon stating that the greater the deviation of a random variate from its mean, the greater the probability that the next measured variate will deviate less far. In other words, an extreme event is likely to be followed by a less extreme event.”

It can be seen in the Fig. 1.1 & Fig. 1.2 that, the further we move
away from the mean returns, the value of probability distribution
function decreases, making the probability of having further higher
deviation in returns less probable.

This might not be the best example to explain mean reversion but here we can see the tendency of periodic returns to return towards the mean which
can explain the validity of mean reversion of freely traded instruments.

How does mean reversion compare with trend following strategies?

An attempt at building my own Mean Reversion Strategy for Crypto Currencies:

Disclaimer : This resource is for educational purpose
only and does not constitute a recommendation to buy, sell or otherwise
deal in investments or trading.

Entry Condition : SELL when RSI MA(14,close,triple,2,0) crosses below RSI MA(14,close,simple,10,0) and 14 RSI higher than 70.

Exit Condition : BUY at stop loss of 1.0% or target profit of 2.5%.

Cryptos have been in a bullish trend recently. Here we are trying to create a
bearish strategy to demonstrate how corrections in prices can be traded
within a broader general trend that is against our direction of trade by
trying to catch the mean reverting moves.

RSI higher than 70 condition is used to check for overbought levels. Since RSI itself is a very volatile indicator, we smoothen its value by applying averages.

Then we check for crossover of the smoothened RSI values when the RSI is
already in the overbought region. Crossover of a smaller period MA with a
larger period MA is checked for fading short term relative strength.
And when the setup is complete, we go short.

Instead of defining mean and taking profits at the mean, to keep the strategy simple, profit taking is done on fixed %. The strategy was backtested to evaluate the performance. While it does not work for all the pairs, it performs well on some other pairs. Here are some of the successful backtest results :

It can be seen in the above images that against a positive period return of
16.26%, the short strategy has managed to make 7% returns by doing 4
trades out of which 3 were profitable. The backtest period is 03/17/2019
to 04/16/2019 and it is carried out on 10 min candle interval.

Conclusion

While there are evidences mentioned in the article suggest that mean
reversion works in freely traded markets, there is a lot of room for
further research. Simple strategies can be created to trade counter to
the direction of trend and also for the periods of sideways trends.

Traders must consider to diversify their strategies across trend
following and mean reversion which might be helpful in reducing
drawdowns or risks associated with using just one kind of strategy.

Bibliography :

https://streak.world for backtesting strategies and charting.

http://mathworld.wolfram.com/ReversiontotheMean.html for defining mean reversion.

The data used to calculate the probability distribution of price changes can be downloaded by clicking on the tickers BTCUSD, BTCUSDC.